Starter · Alpha

Forecast modeling

Sweep hyperparameters on Arc Research feature exports, compare walk-forward results to a naive baseline, publish your own models, and let admins promote winners to the platform core forecasts your research pages use.

Get Arc Research Starter Read the docs

Starter subscribers: ask Archie to submit a Commodity ML alpha access request, or email [email protected].

How it works

Experiment, publish, promote

Parameter sweeps: Grid-search ridge regularization and feature scaling with chronological walk-forward scoring.
Publish your model: Alpha users publish personal models without replacing the platform’s core forecasts.
Promote to core: Admins promote winning trials so scheduled jobs and forecast pages serve that model.

MCP tools for agents →